Abstract

Multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) has been considered as a strong candidate for the beyond 3G (B3G) wireless communication systems, due to its high data-rate wireless transmission performance. It is well known that the advantages promised by MIMO-OFDM systems rely on the precise knowledge of the channel state information (CSI). In real wireless environments, however, the channel condition is unknown. Therefore, channel estimation is of crucial importance in MIMO-OFDM systems. Semi-blind channel estimation as a combination of the training-based or pilot-assisted method and the pure blind approach is considered to be a feasible solution for practical wireless systems due to its better estimation accuracy as well as spectral efficiency. In this thesis, we address the semi-blind channel estimation issue of MIMO-OFDM systems with an objective to develop very efficient channel estimation approaches. In the first part of the dissertation, several nulling-based semi-blind approaches are presented for the estimation of time-domain MIMO-OFDM channels. By incorporating a blind constraint that is derived from MIMO linear prediction (LP) into a training-based least-square method, a semi-blind solution for the time-domain channel estimation is first obtained. It is revealed through a perturbation analysis that the semi-blind solution is not subject to signal perturbation and therefore is superior to pure blind estimation methods. The LP-based semi-blind method is then extended for the channel estimation of MIMO-OFDM systems with pulse-shaping. By exploiting the pulse-shaping filter in the transmitter and the matched filter in the receiver, a very efficient semi-blind approach is developed for the estimation of sampling duration based multipath channels. A frequency-domain correlation matrix estimation algorithm is also presented to facilitate the computation of time-domain second-order statistics required in the LP-based method. The nulling-based semi-blind estimation issue of sparse MIMO-OFDM channels is also addressed. By disclosing and using a relationship between the positions of the most significant taps (MST) of the sparse channel and the lags of nonzero correlation matrices of the received signal, a novel estimation approach consisting of the MST detection and the sparse channel estimation, both in a semi-blind fashion, is developed. An intensive simulation study of all the proposed nulling-based methods with comparison to some existing techniques is conducted, showing a significant superiority of the new methodologies. The second part of the dissertation is dedicated to the development of two signal-perturbation-free (SPF) semi-blind channel estimation algorithms based on a novel transmit scheme that bears partial information of the second-order statistics of the transmitted signal to receiver. It is proved that the new transmit scheme can completely cancel the signal perturbation error in the noise-free case, thereby improving largely the estimation accuracy of correlation matrix for channel estimation in noisy conditions. It is also shown that the overhead caused by the transmission of the 8PF data is negligible as compared to that of regular pilot signals. By using the proposed transmit scheme, a whitening rotation (WR)-based algorithm is first developed for frequency-domain MIMO-OFDM channel estimation. It is shown through both theoretical analysis and simulation study that the new WR-based algorithm significantly outperforms the conventional WR-based method and the nulling-based semi-blind method. By using MIMO linear prediction, the new WR-based algorithm utilizing the 8PF transmit scheme is then extended for time-domain MIMO-OFDM channel estimation. Computer simulations show that the proposed signal-perturbation-free LP-based semi-blind solution performs much better than the LP semi-blind method without using the proposed transmit scheme, the LS method as well as the nulling-based semi-blind method in terms of the MSE of the channel estimate.